Improve checking tx with txsMap for fixing the inconsistency between mem.txs and mem.txsMap

mempool/clist_mempool.go

@@ -297,6 +297,11 @@ func (mem *CListMempool) checkTxAsync(tx types.Tx, txInfo TxInfo, prepareCb func
 
 // CONTRACT: `caller` should held `mem.updateMtx.RLock()`
 func (mem *CListMempool) prepareCheckTx(tx types.Tx, txInfo TxInfo) error {
+	// For keeping the consistency between `mem.txs` and `mem.txsMap`
+	if _, ok := mem.txsMap.Load(TxKey(tx)); ok {
+		return ErrTxInMap
+	}
+
 	txSize := len(tx)
 
 	if err := mem.isFull(txSize); err != nil {

mempool/clist_mempool_system_test.go

@@ -0,0 +1,227 @@
+package mempool
+
+import (
+	"context"
+	"fmt"
+	"math/rand"
+	"os"
+	"strconv"
+	"strings"
+	"sync"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+
+	"github.com/line/ostracon/abci/example/counter"
+	abci "github.com/line/ostracon/abci/types"
+	"github.com/line/ostracon/config"
+	"github.com/line/ostracon/libs/log"
+	"github.com/line/ostracon/proxy"
+	"github.com/line/ostracon/types"
+	"github.com/stretchr/testify/require"
+)
+
+func setupCListMempool(ctx context.Context, t testing.TB,
+	height int64, size, cacheSize int) *CListMempool {
+	t.Helper()
+
+	var cancel context.CancelFunc
+	_, cancel = context.WithCancel(ctx)
+
+	cfg := config.ResetTestRoot(strings.ReplaceAll(t.Name(), "/", "|"))
+	cfg.Mempool = config.DefaultMempoolConfig()
+	logLevel, _ := log.AllowLevel("info")
+	logger := log.NewFilter(log.NewOCLogger(log.NewSyncWriter(os.Stdout)), logLevel)
+
+	appConn := proxy.NewAppConns(proxy.NewLocalClientCreator(counter.NewApplication(false)))
+	require.NoError(t, appConn.Start())
+
+	t.Cleanup(func() {
+		os.RemoveAll(cfg.RootDir)
+		cancel()
+		appConn.Stop() // nolint: errcheck // ignore
+	})
+
+	if size > -1 {
+		cfg.Mempool.Size = size
+	}
+	if cacheSize > -1 {
+		cfg.Mempool.CacheSize = cacheSize
+	}
+	mem := NewCListMempool(cfg.Mempool, appConn.Mempool(), height)
+	mem.SetLogger(logger)
+	return mem
+}
+
+func TestCListMempool_SystemTestWithCacheSizeDefault(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	mem := setupCListMempool(ctx, t, 1, -1, -1) // size=5000, cacheSize=10000
+	recvTxCnt := &receiveTxCounter{}
+	stop := make(chan struct{}, 1)
+	go gossipRoutine(ctx, t, mem, recvTxCnt, stop)
+	makeBlocksAndCommits(ctx, t, mem)
+	close(stop)
+
+	// check the inconsistency
+	require.Equal(t, mem.txs.Len(), sizeOfSyncMap(&mem.txsMap))
+
+	expected := int64(0)
+	actual := recvTxCnt.threadSafeCopy()
+	assert.NotEqual(t, expected, actual.sent, fmt.Sprintf("actual %d", actual.sent))
+	assert.NotEqual(t, expected, actual.success, fmt.Sprintf("actual %d", actual.success))
+	assert.NotEqual(t, expected, actual.failInMap, fmt.Sprintf("actual %d", actual.failInMap))
+	assert.NotEqual(t, expected, actual.failInCache, fmt.Sprintf("actual %d", actual.failInCache))
+	assert.Equal(t, expected, actual.failTooLarge)
+	assert.NotEqual(t, expected, actual.failIsFull, fmt.Sprintf("actual %d", actual.failIsFull))
+	assert.Equal(t, expected, actual.failPreCheck)
+	assert.Equal(t, expected, actual.abciFail)
+}
+
+func sizeOfSyncMap(m *sync.Map) int {
+	length := 0
+	m.Range(func(_, _ interface{}) bool {
+		length++
+		return true
+	})
+	return length
+}
+
+func createProposalBlockAndDeliverTxs(
+	mem *CListMempool, height int64) (*types.Block, []*abci.ResponseDeliverTx) {
+	// mempool.lock/unlock in ReapMaxBytesMaxGasMaxTxs
+	txs := mem.ReapMaxBytesMaxGasMaxTxs(mem.config.MaxTxsBytes, 0, int64(mem.config.Size))
+	block := types.MakeBlock(height, txs, nil, nil)
+	deliverTxResponses := make([]*abci.ResponseDeliverTx, len(block.Txs))
+	for i, tx := range block.Txs {
+		deliverTxResponses[i] = &abci.ResponseDeliverTx{
+			Code: abci.CodeTypeOK,
+			Data: tx,
+		}
+	}
+	return block, deliverTxResponses
+}
+
+func commitBlock(ctx context.Context, t *testing.T,
+	mem *CListMempool, block *types.Block, deliverTxResponses []*abci.ResponseDeliverTx) {
+	mem.Lock()
+	defer mem.Unlock()
+	err := mem.Update(block, deliverTxResponses, nil)
+	require.NoError(t, err)
+}
+
+func receiveTx(ctx context.Context, t *testing.T,
+	mem *CListMempool, tx []byte, receiveTxCounter *receiveTxCounter) {
+	atomic.AddInt64(&receiveTxCounter.sent, 1)
+	txInfo := TxInfo{}
+	// mempool.lock/unlock in CheckTxAsync
+	mem.CheckTxAsync(tx, txInfo,
+		func(err error) {
+			if err != nil {
+				switch err {
+				case ErrTxInCache:
+					atomic.AddInt64(&receiveTxCounter.failInCache, 1)
+				case ErrTxInMap:
+					atomic.AddInt64(&receiveTxCounter.failInMap, 1)
+				}
+				switch err.(type) {
+				case ErrTxTooLarge:
+					atomic.AddInt64(&receiveTxCounter.failTooLarge, 1)
+				case ErrMempoolIsFull:
+					atomic.AddInt64(&receiveTxCounter.failIsFull, 1)
+				case ErrPreCheck:
+					atomic.AddInt64(&receiveTxCounter.failPreCheck, 1)
+				}
+			}
+		},
+		func(res *abci.Response) {
+			resCheckTx := res.GetCheckTx()
+			if resCheckTx.Code != abci.CodeTypeOK && len(resCheckTx.Log) != 0 {
+				atomic.AddInt64(&receiveTxCounter.abciFail, 1)
+			} else {
+				atomic.AddInt64(&receiveTxCounter.success, 1)
+			}
+		})
+}
+
+type receiveTxCounter struct {
+	sent         int64
+	success      int64
+	failInMap    int64
+	failInCache  int64
+	failTooLarge int64
+	failIsFull   int64
+	failPreCheck int64
+	abciFail     int64
+}
+
+func (r *receiveTxCounter) threadSafeCopy() receiveTxCounter {
+	return receiveTxCounter{
+		sent:         atomic.LoadInt64(&r.sent),
+		success:      atomic.LoadInt64(&r.success),
+		failInMap:    atomic.LoadInt64(&r.failInMap),
+		failInCache:  atomic.LoadInt64(&r.failInCache),
+		failTooLarge: atomic.LoadInt64(&r.failTooLarge),
+		failIsFull:   atomic.LoadInt64(&r.failIsFull),
+		failPreCheck: atomic.LoadInt64(&r.failPreCheck),
+		abciFail:     atomic.LoadInt64(&r.abciFail),
+	}
+}
+
+func gossipRoutine(ctx context.Context, t *testing.T, mem *CListMempool,
+	receiveTxCounter *receiveTxCounter, stop chan struct{}) {
+	for i := 0; i < nodeNum; i++ {
+		select {
+		case <-stop:
+			return
+		default:
+			go receiveRoutine(ctx, t, mem, receiveTxCounter, stop)
+		}
+	}
+}
+
+func receiveRoutine(ctx context.Context, t *testing.T, mem *CListMempool,
+	receiveTxCounter *receiveTxCounter, stop chan struct{}) {
+	for {
+		select {
+		case <-stop:
+			return
+		default:
+			tx := []byte(strconv.Itoa(rand.Intn(mem.config.CacheSize * 2)))
+			// mempool.lock/unlock in CheckTxAsync
+			receiveTx(ctx, t, mem, tx, receiveTxCounter)
+			if receiveTxCounter.sent%2000 == 0 {
+				time.Sleep(time.Second) // for avoiding mempool full
+			}
+		}
+	}
+}
+
+func makeBlocksAndCommits(ctx context.Context, t *testing.T, mem *CListMempool) {
+	for i := 0; i < blockNum; i++ {
+		block, deliverTxResponses := createProposalBlockAndDeliverTxs(mem, int64(i+1))
+		time.Sleep(randQuadraticCurveInterval(deliveredTimeMin, deliveredTimeMax, deliveredTimeRadix))
+		commitBlock(ctx, t, mem, block, deliverTxResponses)
+		time.Sleep(randQuadraticCurveInterval(blockIntervalMin, blockIntervalMax, blockIntervalRadix))
+	}
+}
+
+const (
+	nodeNum            = 1
+	blockNum           = 10
+	blockIntervalMin   = 1.0 // second
+	blockIntervalMax   = 1.0 // second
+	blockIntervalRadix = 0.1
+	deliveredTimeMin   = 2.0  // second
+	deliveredTimeMax   = 10.0 // second
+	deliveredTimeRadix = 0.1
+)
+
+func randQuadraticCurveInterval(min, max, radix float64) time.Duration {
+	rand.Seed(time.Now().UnixNano())
+	x := rand.Float64()*(max-min) + min
+	y := (x * x) * radix
+	return time.Duration(y*1000) * time.Millisecond
+}

mempool/errors.go

@@ -6,7 +6,9 @@ import (
 )
 
 var (
-	// ErrTxInCache is returned to the client if we saw tx earlier
+	// ErrTxInMap is returned to the client if we saw tx earlier in txsMap
+	ErrTxInMap = errors.New("tx already exists in txsMap")
+	// ErrTxInCache is returned to the client if we saw tx earlier in cache
 	ErrTxInCache = errors.New("tx already exists in cache")
 )